Forgery Detection and Authentication of Medical Images using Adaptive Bit Substitution Watermarking

The primary objective of this research is to develop an adaptive digital watermarking framework specifically designed for medical image authentication and forgery detection. The methodology employs a systematic workflow involving RGB medical image acquisition, pre-processing, and the division of images into non-overlapping 4×4 blocks. The core techniques integrate adaptive bit substitution, keyed pixel selection, and cryptographic hash-based key generation derived from patient or camera identification data. Unlike conventional methods, the proposed algorithm dynamically selects the optimal embedding bit-plane—LSB, Bit2SB, or Bit3SB—based on local image characteristics. The proposed model was rigorously evaluated through a comparative analysis against conventional fixed-bit substitution techniques, including Single-LSB, Bit2SB, and Bit3SB. Main findings demonstrate that the adaptive approach significantly outperforms fixed schemes across multiple metrics, including MAE, MSE, SSIM, UIQI, and PSNR. Specifically, the system achieved PSNR values exceeding 61 dB and maintained a reliable forgery detection rate with correlation values consistently above 0.99. This study contributes a unified solution that balances imperceptibility and security, offering practical benefits for digital healthcare infrastructures by preserving diagnostic quality while providing robust resistance to intentional tampering.